Pioneering Agentic Payments Technology to Power Commerce in the Age of AI

Digital commerce is shifting from clicks to outcomes. For thirty years, people learned the grammar of software – menus, forms, checkouts. Now we are moving from manual, command-based interactions to a world in which people and businesses delegate a range of objectives to intelligent systems that can reason, plan and act on their behalf. That demands a new operating model where trust, policy and payments are built in – not bolted on. The enabling technology is still emerging, but the transformation it heralds is so profound that forward-thinking businesses must act now to stay ahead.

Artificial intelligence (AI) is changing commerce as we know it, with consumer behavior rapidly shifting from traditional search engines to AI for product discovery and purchases. Today, ChatGPT has around 700 million weekly active users, a four-fold year-over-year increase.^[1] Against this background, over half (58%) of consumers have replaced traditional search engines with Generative AI tools for product/service recommendations, up significantly from 25% in 2023.^[2] In February 2025, traffic to U.S. retail websites from Generative AI sources jumped 1,200% compared to July 2024,^[3] with Agentic AI expected to handle up to 20% of eCommerce tasks in 2025.^[4]

An AI agent is any system that utilizes artificial intelligence to perform tasks on a user’s behalf. AI agents come in a variety of forms, including chatbots, voice assistants, in-vehicle software and even smart appliances (such as a refrigerator that automatically orders groceries each week). Agentic use cases are expanding rapidly, with eCommerce being one of the first to attract users. Agentic commerce (aCommerce) takes conversational AI platforms further by enabling AI agents to perform commerce related tasks for the user – searching for products, comparing prices, adding things to a cart and completing checkout within clear guardrails using the user’s stored payment credentials:

• Interpret a goal: The user begins by expressing a desired outcome in natural language, such as “Buy me the best running shoes under $100.” The agentic system is designed to understand this intent, parsing not just the literal request but also any underlying preferences, constraints or context – like preferred brands, delivery dates and budget limits.
• Plan and use tools: Once the goal is understood, the system intelligently breaks down the objective into actionable steps. It determines what needs to happen – such as searching for suitable products, comparing prices, checking inventory and reviewing return policies. To accomplish these tasks, the agent leverages a range of digital tools – including apps, web browsers and APIs – and seamlessly orchestrates them to build an optimal path to the user’s goal.
• Execute within guardrails: The agentic system carries out the plan within clearly defined boundaries. Execution is governed by delegation of the user’s authority ensuring transparency and control. The system also operates within pre-set policies, such as spending caps or approval requirements, so users retain oversight and can intervene or revoke permissions at any time.

This means consumers can let AI assistants handle routine tasks like reordering or solving problems while they focus on important decisions.

As AI agents autonomously conduct payments on behalf of users, they introduce new challenges to payment flows:

In practical terms, what does this mean for payments, and how must the payments ecosystem adapt to embrace this change?

In agentic commerce, users delegate purchasing decisions to AI agents. The user no longer visits the merchant’s website or engages directly with their brand. Instead, the agent selects and transacts on the user’s behalf – often without the user even knowing which merchant fulfilled the order. Consequently, merchants lose visibility and control over customer experience, weakening brand recognition and loyalty, and in some cases might not even know the purchase was an agentic purchase (since some AI agents today are simply form fill injecting a user’s payment details into the web checkout).

On the other hand, if an agent makes a purchasing decision that leads to user dissatisfaction, it may be unclear who is liable – merchant, user, issuer or agent. Managing returns, disputes and customer service becomes more complex in an AI-mediated transaction. For example, if the instruction to an AI agent was to “Buy the best size 12 running shoes” and the AI agent mistakenly buys 12 pairs of shoes (not size 12), that raises the question of who is responsible when an AI agent makes a wrong purchase. Unlike humans, who might only make a handful of mistaken purchases per year, agents could generate thousands of transactions, and even a 1% error rate could mean massive dispute volumes bringing a whole new ecosystem problem for merchants and issuers.

Historically, payment networks have been built on the foundational assumption that a human being is the initiator of every purchase. Fraud detection systems are trained to recognize typical human behaviors – such as time of day, location, device type and purchase cadence. These systems flag deviations from those patterns as potential fraud. However, in aCommerce, the agent’s behavior may appear unusual or high-risk to existing fraud systems. For example, an agent might transact at odd hours, across geographies, or perform rapid, repeated purchases – all of which could mimic the behavior of a fraud bot. As a result, even authorized transactions by trusted AI agents may be incorrectly declined, eroding the customer experience and undermining trust in the system.

In summary, the use of agents in aCommerce surfaces new issues to the evolving payment ecosystem, namely:

• Customer intent: how to ensure AI actions are aligned with users’ intent?
• Fraud & declines: how to adapt fraud detection and prevention models that distinguish between legitimate agent behavior and true fraud, as well as new identifiers to recognize and authenticate AI-initiated transactions in real time?
• Disputes: how to address the dispute process, such as chargeback processes and arbitration mechanisms?

Mastercard is working with Microsoft, and other companies including AI platforms, to integrate Microsoft’s leading AI technologies with Mastercard’s trusted payment solutions to develop and scale aCommerce, addressing the evolving needs of the entire commerce value chain. Grounded by the company's commitment to responsible AI, earlier this year we introduced our Agentic Payments Program to ensure that payments being made within AI platforms are safe and transparent at every stage of the transaction – before, during and after. The program introduces Mastercard Agentic Tokens, which build upon our proven tokenization capabilities that today power global commerce solutions.^[5] Tokenization allows AI agents to be uniquely identified and linked to individual users, ensuring that payment credentials are protected and transactions can proceed without repeated user intervention.^[6]

Mastercard is helping to solve the challenges that AI agents introduce in payment ecosystems:

1. Securely registering and authenticating trusted agents: The Agentic Payments Program will require trusted AI agents to be registered and verified, after which they will be able to make secure payments on behalf of their users.^[7]
2. Establishing clear rules for individual control: Users will have complete control over what the agent is allowed to purchase on their behalf, ensuring that the payments they make are authorized and identified.
3. Facilitating safe and secure transactions: Enhanced tokenization technology will enable payments to be initiated through conversational interfaces and conducted at millions of merchants of all sizes supporting online commerce today. Every player in the value chain, from consumers to issuers and merchants, will be able to recognize the transactions that are facilitated by intelligent agents, delivering greater transparency.
4. Protecting against fraud and avoiding unnecessary disputes: Mastercard’s best-in-class cyber, security and authentication capabilities will protect merchants and individuals against bad actors from end-to-end. We are exploring the use of AI agents to facilitate strong consumer authentication leveraging on-device biometrics and a process to help clarify agentic transactions that may be unfamiliar or unrecognized.
5. Developing a verifiable credential standard for payments that confirms payment details such as amount, merchant and product, together with the FIDO Alliance and other industry leaders. This ensures that everyone involved in a transaction can have confidence that it was approved by the user, paving the way for a more secure, seamless and trusted foundation for agentic payments. We are also working with AI and commerce leaders including Stripe, Google and Ant International’s Antom to make secure agentic transactions accessible and scalable for digital merchants and platforms globally.^[8]

We also joined Google and other partners to develop an open protocol designed to serve as the foundation needed to enable AI agents to transact payments on behalf of users and merchants. Since aCommerce breaks the assumption that payments are initiated by humans, a new protocol was needed providing the basis to prove that a user has authorized an agent to make a purchase; to enable a merchant to authenticate that the agent's request reflects the user's intent; and to determine accountability if a fraud or error occurs. The protocol uses Mandates (tamper-proof, cryptographically signed digital contracts) that serve as verifiable proof of a user's instructions. The protocol is designed to be an open, shared protocol that provides a common language for secure, compliant transactions between agents and merchants, helping to prevent a fragmented ecosystem.^[9]

Mastercard is working with partners across the ecosystem^[10] to build the standards and tools that will define aCommerce. We are building the infrastructure for a new generation of intelligent transactions, where consumers and developers can empower AI agents to act on their behalf with trust, transparency and precision.^[11]

About Mastercard

Mastercard powers economies and empowers people in 200+ countries and territories worldwide. Together with our customers, we’re building a sustainable economy where everyone can prosper. We support a wide range of digital payments choices, making transactions secure, simple, smart and accessible. Our technology and innovation, partnerships and networks combine to deliver a unique set of products and services that help people, businesses and governments realize their greatest potential.

Download your sample issue here!

No part of Central Bank Payments News may be reproduced, copied, republished, or distributed in any form or by any means, in whole or in part, without the express and prior written permission of the publisher, Currency Research Malta, Ltd.